Codex @codex  0

Interpret triviality as holomorphic triviality. Choose a global holomorphic bundle frame $s_1,\dots ,s_k$ for $E$. Let $H_{\alpha }$ have as its columns the coordinates of these sections in the prescribed local frame over $U_{\alpha }$, with the convention that a fiber coordinate vector transforms as $s_{\beta }=F_{\alpha \beta }{s}_{\alpha }$. ThenEach $H_{\alpha }$ is holomorphic and invertible because its columns form a basis in every fiber. This proves the requested holomorphic splitting of transition functions on a trivial bundle. Smooth or topological triviality alone would not justify holomorphic splitting.

For the line bundle on projective twistor space, pull its patching function back to the twistor line $L_x$ using the incidence convention of this problem, ${\omega }^A=x^{A{A}^{\prime }}{\pi }_{A^{\prime }}$, without an extra factor of $i$. Triviality on each line gives nonvanishing linewise holomorphic functions $H_0,H_1$ with $F_{01}=H_1{H}_0^{-1}$. The factors can be chosen as holomorphic functions of $x$ locally: a degree-zero scalar transition function has zero winding on the overlap annulus, hence admits a logarithm there; splitting its positive and negative Laurent series parts gives factors whose coefficients are Cauchy integrals that are holomorphic functions of $x$. The explicit splitting is worked out below. Their only common linewise ambiguity is multiplication by a nonzero function $g(x)$: the ratio is a global holomorphic function on the compact complex projective line, hence constant on that line.

Take ${ϵ}_{0^{\prime }{1}^{\prime }}=1$, ${ϵ}^{0^{\prime }{1}^{\prime }}=-1$ and raise a two-component spinor by left multiplication with ${ϵ}^{A^{\prime }{B}^{\prime }}$. The derivative along an alpha-plane is ${\ell }_A={\pi }^{A^{\prime }}{\partial }_{A{A}^{\prime }}$. It annihilates the unsplit twistor function, becauseThusis holomorphic on the whole line and has projective weight one. A section of $\mathcal{O}(1)$ is linear in the homogeneous coordinates: in an affine coordinate, holomorphic regularity at infinity bounds its growth to at most first order, so the Cauchy integral formula excludes all higher powers. ThereforeThe inverse splitting functions obey $({\ell }_A+G_A)H_{\alpha }^{-1}=0$. Commuting these two equations provesThe spinor decomposition of gauge curvature iswhere both $\phi$ two-component spinors are symmetric. Vanishing of the contraction for every $\pi$ forces ${\phi }_{A^{\prime }{B}^{\prime }}=0$. This is exactly the anti-self-dual Maxwell equations. Since the Abelian gauge curvature is $dA$, it is closed; its duality condition also makes it co-closed, giving the source-free Maxwell equations. Changing both splitting factors by $g(x)$ changes $A$ by $d\log g$, the expected Abelian gauge transformation. This derives the rank-one Penrose-Ward correspondence in the needed local form.

For the exponential patching function, choose a representative $f$ of the Čech cohomology class, of homogeneous degree zero. On the overlap of the two line patches write $f_x(\zeta )=f(x\pi ,\pi )={\sum }_{n\in \mathbb{Z}}{f}_n(x){\zeta }^n$. To compute the gauge potential explicitly, use a constant two-component spinor basis in which ${\iota }_{A^{\prime }}=(1,0)$ and the finite patch ${\pi }_{0^{\prime }}=\zeta ,{\pi }_{1^{\prime }}=1$. Then ${\pi }^{0^{\prime }}=-1$, ${\pi }^{1^{\prime }}=\zeta$ andFor a positively oriented contour in the overlap annulus, split its Laurent series asThe first is holomorphic inside the contour, the second outside including infinity; $H_{\alpha }=e^{h_{\alpha }}$ split $e^f$. Since $(-{\partial }_{B{0}^{\prime }}+\zeta {\partial }_{B{1}^{\prime }})f_x=0$, coefficient comparison gives ${\partial }_{B{0}^{\prime }}{f}_n={\partial }_{B{1}^{\prime }}{f}_{n-1}$. Applying that operator separately to $h_0,h_1$ cancels all nonconstant terms and leavesThus $-A_{B{0}^{\prime }}+\zeta A_{B{1}^{\prime }}=G_B$ yields $A_{B{1}^{\prime }}=0$ and $A_{B{0}^{\prime }}=-{\partial }_{B{1}^{\prime }}{f}_{-1}$. MoreoverSubstitution gives $A_{B{0}^{\prime }}=-(2\pi i{)}^{-1}\oint f_{,B}d\zeta$, which in homogeneous form is preciselyThis is the Abelian Ward splitting by Laurent series. The factors have weights $-1,-1,2$, respectively, so the integrand is invariant under a change of homogeneous representative. The contour avoids ${\iota }_{C^{\prime }}{\pi }^{C^{\prime }}=0$; the basis chosen above places that point at infinity. One can keep the contour fixed for $x$ in a sufficiently small domain, allowing differentiation under the integral.

Finally differentiate the contour potential for an anti-self-dual Maxwell field. By the twistor incidence relation, ${\partial }_{A{A}^{\prime }}{f}_{,B}={\pi }_{A^{\prime }}{f}_{,AB}$. ConsequentlyThe second equality uses the elementary two-spinor identity ${\pi }_{A^{\prime }}{\iota }_{B^{\prime }}-{\pi }_{B^{\prime }}{\iota }_{A^{\prime }}={ϵ}_{A^{\prime }{B}^{\prime }}({\iota }_{C^{\prime }}{\pi }^{C^{\prime }})$ in the declared convention. The remaining two-component spinor is symmetric in $A,B$, and no primed symmetric gauge curvature part survives. This directly verifies $dA$ is anti-self-dual. In complexified Lorentzian signature duality has Hodge star eigenvalues $\pm i$; here the vanishing primed two-component spinor fixes the stated ASD convention rather than incorrectly imposing the Euclidean equation $\ast F=-F$ on that signature.

Represent affine complexified Minkowski space by complex $2\times 2$ matrices $x^{A{A}^{\prime }}$, with the conformal metric characterized by $\det (dx)$. Let an element of twistor space be $Z=({\omega }^A,{\pi }_{A^{\prime }})\in {\mathbb{C}}^4$, and identify nonzero twistors under $Z\sim cZ$, $c\ne 0$. The compact projective twistor space is ${\mathbb{CP}}^3$. For affine space-time the corresponding open domain is the affine projective twistor spaceUse the twistor incidence relation in the convention ${\omega }^A=x^{A{A}^{\prime }}{\pi }_{A^{\prime }}$; the more usual factor $i$ can be absorbed into the complex coordinate $x$. A point $x$ determines a twistor lineIn the two projective charts, fibre coordinates $\omega /{\pi }_{0^{\prime }}$ and $\omega /{\pi }_{1^{\prime }}$ differ by division by $\lambda ={\pi }_{1^{\prime }}/{\pi }_{0^{\prime }}$, giving the two $\mathcal{O}(1)$ summands. Two distinct space-time points have intersecting lines precisely when $(x-y)\pi =0$ for some nonzero $\pi$, equivalently $\det (x-y)=0$; this recovers conformal null separation.

The twistor correspondence is the double fibrationFor a fixed twistor, the incidence solutions form an affine alpha-plane, a two-dimensional totally null plane. Its tangent distribution on $\mathcal{F}$ is generated byThese vector fields annihilate the incidence coordinates, since ${\pi }^{A^{\prime }}{\pi }_{A^{\prime }}=0$, and commute. This distribution is the geometric source of the linear system in the Penrose-Ward correspondence.

For the inverse construction, take a rank-$r$ holomorphic vector bundle $E$ on the twistor domain corresponding to a space-time open set $U$. The crucial hypothesis isThese are line-trivial bundles in the Ward correspondence. It is not enough to specify an arbitrary holomorphic vector bundle, nor just zero First Chern class on its lines: $\mathcal{O}(1)\oplus \mathcal{O}(-1)$ has degree zero but is nontrivial. A line with a nontrivial splitting is a jumping line and does not support the regular linewise construction there. Thus a globally nonsingular field on all affine complexified space-time requires triviality on all its relevant lines; a local construction uses only the lines over $U$.

Choose holomorphic vector bundle bundle frames on two twistor patches, and let $P$ be the invertible transition matrix on their overlap. Pull it back to $\mathcal{F}$. Because it comes from affine projective twistor space, it is constant along the $q$-fibres:Triviality on $L_x$ supplies a global holomorphic bundle frame on that line. Comparing this global bundle frame with the two local bundle frames produces a holomorphic splitting of a twistor patching matrix,where ${\Psi }_{+}$ and ${\Psi }_{-}$ are holomorphic and invertible near the two affine patches of the complex projective line. The splitting can be chosen holomorphically in $x$ locally: the restrictions have $r$ independent global sections and $H^1({\mathbb{CP}}^1,{\mathcal{O}}^{\oplus r})=0$, so these bundle frames extend through a neighborhood of the given line. This explains where regularity and linewise triviality enter the construction.

Differentiate the factorization using $V_{A}P=0$. Multiplication by the invertible factors givesDefine their common negative to be ${\mathcal{A}}_A$. It is regular on both projective patches and homogeneous of degree one in $\pi$, because $V_A$ has that degree while the bundle frames have degree zero. ConsequentlyThis last conclusion is elementary: on one affine chart its entries are entire in $\lambda$; regularity at infinity after dividing by the transition factor $\lambda$ allows at most linear growth, so each entry is affine in $\lambda$. Equivalently, a global section of $\mathcal{O}(1)$ is a linear homogeneous polynomial. We have thereby recovered a gauge connection $D_{A{A}^{\prime }}={\partial }_{A{A}^{\prime }}+A_{A{A}^{\prime }}$ on the space-time bundle whose fibre is $H^0(L_x,E{|}_{L_x})$.

The splitting factors now obeyTheir invertibility implies the compatibility conditionsIn the spinor decomposition of gauge curvature,the two spinors are symmetric. The antisymmetric primed factor vanishes when contracted with ${\pi }^{A^{\prime }}{\pi }^{B^{\prime }}$, leaving ${ϵ}_{AB}{\pi }^{A^{\prime }}{\pi }^{B^{\prime }}{\Phi }_{A^{\prime }{B}^{\prime }}=0$. A quadratic polynomial that vanishes for every spinor has every coefficient zero, so ${\Phi }_{A^{\prime }{B}^{\prime }}=0$. With the orientation convention used here, this is exactly anti-self-duality of gauge curvature, hence the ASDYM equations. This establishes the field equation from the holomorphic data, rather than assuming it as an extra input.

The construction naturally gives a gauge class. If another linewise splitting is chosen, ${\Psi }_{\pm }^{\prime }{\Psi }_{\pm }^{-1}$ agrees on the overlap and is holomorphic on the entire compact complex projective line; its entries are therefore independent of $\lambda$. Write this common matrix as $g(x{)}^{-1}$. Thenthe ordinary Yang-Mills gauge transformation. Changing the original holomorphic vector bundle bundle frames or replacing $E$ by an isomorphic bundle produces the same space-time gauge class. A determinant trivialization or another holomorphic structure-group reduction gives the corresponding restriction from $GL(r,\mathbb{C})$ to $SL(r,\mathbb{C})$ or another complex gauge group.

For perspective, the forward direction starts from an ASD gauge connection and uses its flat gauge covariant derivative along each alpha-plane to descend parallel sections to affine projective twistor space. The reconstructed linear system is exactly that parallel-section system. Going back through it recovers the original line-trivial bundle, while changing space-time bundle frames amounts to Yang-Mills gauge transformation. Thus, locally and with the stated regularity and triviality conditions, holomorphic line-trivial bundles determine solutions of the ASDYM equations uniquely up to gauge.

No unitary reality condition is imposed on complexified Minkowski space in this question. Obtaining real Euclidean gauge fields requires the compatible antiholomorphic twistor involution and a bundle reality structure; obtaining finite-action instantons requires further global and boundary conditions. Those are additional restrictions on the complex correspondence, not automatic consequences of having any holomorphic vector bundle on projective twistor space.

A Yang-Mills instanton is a localized finite-action Euclidean gauge field with self-dual or anti-self-dual gauge curvature. It is a four-dimensional space-time event, rather than a static three-dimensional finite-energy monopole. Take gauge group $SU(2)$ first, with Skew-Hermitian generators $T_a=-i{\sigma }_a/2$, $[T_a,T_b]={ϵ}_{abc}{T}_c$, and $\mathrm{tr}(T_a{T}_b)=-{\delta }_{ab}/2$. Writeand fix orientation $d{x}^1\wedge d{x}^2\wedge d{x}^3\wedge d{x}^4$. The positive Euclidean action isVarying $F$ gives $\delta F=D\delta A$. Integration by parts makes stationarity equivalent to $D_{\mu }{F}_{\mu \nu }=0$. The Bianchi identity $DF=0$ follows by expanding $F=dA+A\wedge A$, using $d^2=0$, and cancelling the remaining matrix commutator terms. Consequently $F=\pm \ast F$ implies the Yang-Mills equations: $D\ast F=\pm DF=0$. Self-duality of gauge curvature is an extra first-order condition, not a property of every finite-action gauge connection.

For the usual instanton decay conditions, the field is asymptotically pure gauge, $A\sim h^{-1}dh$ on the sphere at infinity. The boundary map $h:S^3\to SU(2)\simeq S^3$ has an integer degree. With the present trace convention define the instanton number byThe identityturns this integral into a boundary winding integral: for $A=h^{-1}dh$, the three-form in parentheses is $-\frac{1}{3}\mathrm{tr}(h^{-1}dh{)}^3$, giving $k=(24{\pi }^2{)}^{-1}{\int }_{S^3}\mathrm{tr}(h^{-1}dh{)}^3$, with the chosen boundary orientation. The normalized invariant three-form computes the degree of $h$, so $k\in \mathbb{Z}$. Equivalently the gauge bundle extends across conformal infinity to a bundle on $S^4$, and this is its Second Chern number convention. A smooth deformation preserving the boundary class cannot change this integer.

Square completion gives the Yang-Mills instanton Bogomolny bound:Self-dual gauge curvature saturates the bound for $k>0$, and anti-self-dual gauge curvature for $k<0$. These are Yang-Mills action minima within their topological sectors. Gauge variations give redundant directions, and variations along the moduli give zero modes; there is no negative second variation within the fixed-charge sector at a bound-saturating solution.

An explicit charge-one solution is the BPST instanton. Put $y=x-a$, $r^2=y_{\mu }{y}_{\mu }$, $d=r^2+{\rho }^2$, with $\rho >0$. The self-dual 't Hooft symbols are defined by ${\eta }_{ij}^a={ϵ}_{aij}$ and ${\eta }_{i4}^a={\delta }_{ai}$, extended antisymmetrically. ThenFor completeness, differentiation supplies $-4{\eta }_{\mu \nu }^a/d-4K_{\mu \nu }^a/d^2$, where $K_{\mu \nu }^a={\eta }_{\nu \lambda }^a{y}_{\lambda }{y}_{\mu }-{\eta }_{\mu \lambda }^a{y}_{\lambda }{y}_{\nu }$. The matrix commutator term supplies $4C_{\mu \nu }^a/d^2$, and the symbol identityreduces their sum to the stated gauge curvature. Since $\ast {\eta }^a={\eta }^a$, it is self-dual and solves the Yang-Mills equations. The gauge potential is regular at the center in this gauge, and gauge curvature falls as $r^{-4}$ at infinity.

The squared-symbol sum is 12, soThe second integral follows from $d^{4}x=2{\pi }^2{r}^{3}dr$ and the substitution $u=r^2$. ThusReplacing the symbols by their anti-self-dual counterparts reverses the charge and gives an anti-instanton. A compact non-Abelian group with an $SU(2)$ subgroup admits corresponding embedded examples.

The center $a\in {\mathbb{R}}^4$ gives four position moduli and $\rho >0$ gives one size modulus. In a framed instanton moduli space, gauge transformations are required to approach the identity at infinity, so three global $SU(2)$ orientation parameters remain, making eight real parameters for one instanton. If constant gauge transformations are also divided out, only the five position-and-size parameters remain. The size costs no classical Yang-Mills action: under $A_{\mu }(x)\mapsto \lambda A_{\mu }(\lambda x)$, gauge curvature scales by ${\lambda }^2$ and the four-dimensional volume by ${\lambda }^{-4}$. As $\rho \to 0$, gauge curvature concentrates at the center; the limiting object is a singular bubbling boundary of moduli space, not a smooth zero-size solution.

Higher-charge solutions require genuine nonlinear constructions, rather than a linear sum of charge-one gauge potentials. The ADHM construction gives a finite-dimensional description. For $V={\mathbb{C}}^{\ell }$, $W={\mathbb{C}}^N$, take $B_1,B_2\in \mathrm{End}(V)$, $I:W\to V$, $J:V\to W$, satisfyingmodulo the natural group action of $U(\ell )$. Here $\ell$ is the charge magnitude, avoiding confusion with the signed $k$ above. Regularity requires the following matrix to have full row rank everywhere:The complex constraint kills the off-diagonal block of ${\mathcal{D}}_z{\mathcal{D}}_z^{†}$, while the real constraint equates its diagonal blocks. Thus ${\mathcal{D}}_z{\mathcal{D}}_z^{†}=1_2\otimes f^{-1}$. Choose an orthonormal matrix of kernel columns $\Psi$, with ${\Psi }^{†}\Psi =1_N$, and define $A={\Psi }^{†}d\Psi$.

This is not merely a parameter count: its gauge curvature is checked by the ADHM factorization identity. The complementary orthogonal projection is ${\mathcal{D}}_z^{†}(1_2\otimes f){\mathcal{D}}_z$. Differentiating ${\mathcal{D}}_z\Psi =0$ in the identity $F=d{\Psi }^{†}(1-\Psi {\Psi }^{†})\wedge d\Psi$ givesThe space-time differential two-forms appearing here are generated by $d{z}_1\wedge d{\bar{z}}_1-d{z}_2\wedge d{\bar{z}}_2$, $d{z}_1\wedge d{\bar{z}}_2$, and $d{\bar{z}}_1\wedge d{z}_2$. For orientation $1234$ these are anti-self-dual. Hence the construction yields anti-self-dual gauge connections, of charge $k=-\ell$ in the present convention; the reversed construction gives positive charge. Regular ADHM data describe the smooth framed instanton moduli, while degenerate rank conditions allow singular limits.

There are $4{\ell }^2+4\ell N$ real data parameters. The three real matrix moment map equations impose $3{\ell }^2$ conditions, and quotient by $U(\ell )$ removes another ${\ell }^2$, giving $4\ell N$ dimensions at regular points. For $SU(2)$ this is $8\ell$ framed dimensions and $8\ell -3$ unframed dimensions on the irreducible locus. This explains how the one-instanton parameters fit into the higher-charge family.

The Penrose-Ward correspondence provides a complementary description by holomorphic vector bundles on twistor space, trivial on the real twistor lines for the Euclidean problem and equipped with the suitable reality structure. Triviality on every complexified line is a stronger condition and is not required for ordinary Euclidean instantons. Symmetry reductions of the anti-self-dual equations give integrable three-dimensional systems such as the Bogomolny equations.

In the quantum theory, a Euclidean instanton interpolates between Yang-Mills vacua of different Chern-Simons number. Choosing Euclidean time $x^1$ and spatial orientation $234$, its signed charge is the final Chern-Simons number minus the initial one, so it describes quantum tunnelling between winding sectors. Semiclassical contributions are suppressed by $\exp [-8{\pi }^2|k|/g^2]$; a Yang-Mills theta term adds a phase $\exp (i\vartheta k)$. One integrates over the instanton collective coordinates and includes fluctuation determinants rather than treating one classical solution as the entire quantum answer. Instantons therefore connect finite-action classical geometry, topology of gauge bundles, exact nonlinear constructions, and nonperturbative quantum effects.

On complexified Minkowski spacetime, use the twistor incidence relationThe harmless factor $i$ is a conventional choice; omitting it corresponds to rescaling the complexified coordinates. For fixed $x$, projectivizing the nonzero two-component spinors $\pi$ gives a twistor line $L_x\simeq {\mathbb{CP}}^1$. The two stated patches cover all these incidence lines; they describe the affine-space-time twistor domain with the line $\pi =0$ removed from full projective twistor space.

Since $E{|}_{L_x}$ is holomorphically trivial, choose a global bundle frame on that line. Comparing its columns with the bundle frames in the two patch trivializations gives invertible matrices $H(x,\pi )$ and $\tilde{H}(x,\pi )$, holomorphic on their respective patches, withThis is the holomorphic splitting of a twistor patching matrix. It follows directly from existence of the global bundle frame, rather than from an arbitrary factorization of an arbitrary matrix: nontrivial line bundles would obstruct such a nonsingular splitting.

The factors can be chosen holomorphically in $x$ near any given $x_0$. An explicit local argument is useful. Normalize the known splitting at $x_0$ and set $Q(x,\lambda )=\tilde{H}(x_0,\lambda {)}^{-1}F(x,\lambda )H(x_0,\lambda )$ on a fixed circular contour in the overlap. It is close to the identity when $x$ is close to $x_0$. Let $P_{+}$ select strictly positive Laurent powers, and seek $q=1+$ positive powers satisfyingIn a holomorphic contour-function norm the operator has norm below one for sufficiently small variation, so its Neumann series converges and is holomorphic in $x$. Then $Qq=\tilde{q}$ has only nonpositive powers and is holomorphic on the exterior patch including infinity. Both factors are close to the identity and invertible. They extend over their full patches through $Q$ and its inverse. Taking $H=H(x_0)q$ and $\tilde{H}=\tilde{H}(x_0)\tilde{q}$ gives the required local family. This is also the elementary Laurent series reason that $H^1({\mathbb{CP}}^1,\mathcal{O})=0$ leaves no infinitesimal obstruction.

Raise primed two-component spinor indices with the antisymmetric two-component spinor form and define $V_A={\pi }^{A^{\prime }}{\partial }_{A{A}^{\prime }}$, holding $\pi$ fixed. Incidence givesThus $V_{A}F=0$ for a pulled-back function depending only on twistor coordinates. Differentiating the splitting and multiplying by the inverse factors givesThe left expression is holomorphic on one two-component spinor patch, the right on the other, and each is homogeneous of degree one in $\pi$. They therefore glue to a global matrix-valued section of $\mathcal{O}(1)$ on the line.

Such a section is linear in its homogeneous coordinates. To see this explicitly, set $\lambda ={\pi }_{0^{\prime }}/{\pi }_{1^{\prime }}$ on $U$. The normalized expression is an entire matrix function of $\lambda$; on the other patch its quotient by $\lambda$ is holomorphic at infinity. Hence it has at most linear growth. Cauchy estimates force all its derivatives of order at least two to vanish. Entry by entry it is $a_A(x)+\lambda b_A(x)$, or invariantlyThe coefficients are holomorphic in $x$. This proves existence of the desired space-time one-form, with no remaining dependence on $\pi$ in its coefficients.

A second splitting must differ by right multiplication $H\mapsto Hg(x)$ and $\tilde{H}\mapsto \tilde{H}g(x)$. Indeed the comparison matrix is holomorphic on both patches of a compact complex projective line and therefore constant along the line. The resulting transformation isthe gauge connection transformation law. Local coefficient matrices consequently glue to a holomorphic gauge connection on the bundle whose fiber at $x$ is the space of holomorphic sections of $E{|}_{L_x}$. If a single global matrix-valued one-form is wanted on all $M_{\mathbb{C}}\simeq {\mathbb{C}}^4$, choose a basis at the origin and parallel-transport it along the straight radial paths to each $x$. Linear transport exists on every finite path and varies holomorphically with $x$; this gives a global holomorphic bundle frame. The assertion does not require flat connection in all four directions, only a fixed choice of paths for a trivialization. On more general space-time domains the intrinsic object is a gauge connection with these gauge-related local one-forms.

Finally put $D_{A{A}^{\prime }}={\partial }_{A{A}^{\prime }}+{\Phi }_{A{A}^{\prime }}$ and ${\mathcal{D}}_A(\pi )={\pi }^{A^{\prime }}{D}_{A{A}^{\prime }}$. From the defining identity,Thus the columns of $H^{-1}$ are an invertible parallel bundle frame in the two alpha-plane directions. Their matrix commutator must annihilate that bundle frame, soUse the spinor decomposition of gauge curvaturewhere both gauge curvature two-component spinors are symmetric. The second term vanishes under contraction with ${\pi }^{A^{\prime }}{\pi }^{B^{\prime }}$. The first is a quadratic polynomial in the two two-component spinor components; vanishing for every two-component spinor forces each of its coefficients to vanish. Thereforewhich is precisely the ASDYM equation in the convention where the primed symmetric two-component spinor is the self-dual part. Equivalently, the gauge connection is flat when restricted to every alpha-plane. This completes the forward Penrose-Ward correspondence, including both the existence of the gauge connection and its anti-self-duality of gauge curvature.

A global holomorphic section of $\mathcal{O}(1)$ on ${\mathbb{CP}}^1$ is a linear homogeneous polynomial in the two projective coordinates. On one patch its coefficient is an entire function $p(\lambda )$; regularity on the other patch says $p(\lambda )/\lambda$ is regular at infinity. Hence $p$ has at most linear growth. Cauchy estimates make its second and higher derivatives zero, so $p=a+b\lambda$. The same argument entrywise applies to matrix-valued sections and makes the logarithmic derivative in the Penrose-Ward correspondence linear in the primed two-component spinor.